bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[nvp]

[-- Attachment #1.1: Type: text/plain, Size: 1049 bytes --]

Tags: patch


Bug: After `treesit-explorer-mode` is enabled in a buffer and its
associated `treesit--explorer-buffer` is killed, a subsequent call
to `treesit-explorer-mode` initially displays an empty explorer
buffer b/c `treesit--explorer-refresh` sees old value for
`treesit--explorer-last-node`.

* lisp/treesit.el (treesit-explorer-mode): reset
`treesit--explorer-last-node` when `treesit--explorer-buffer` was killed

In GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
3.24.33, cairo version 1.16.0) of 2023-10-05 built on noah-X580VD
Repository revision: 505c80623049d9e181918acdac8229c9a2041b1e
Repository branch: master
Windowing system distributor 'The X.Org Foundation', version 11.0.12101004
System Description: Ubuntu 22.04.3 LTS

Configured using:
 'configure --prefix=/usr/local --with-modules --with-tree-sitter
--with-threads --with-x-toolkit=gtk3 --with-xwidgets --with-gnutls
--with-json --with-mailutils --with-jpeg --with-png --with-rsvg
--with-tiff --with-xml2 --with-xpm --with-imagemagick CC=gcc-12
CXX=gcc-12'

[-- Attachment #1.2: Type: text/html, Size: 1190 bytes --]

[-- Attachment #2: fix-treesit-explorer-last-node.patch --]
[-- Type: text/x-patch, Size: 841 bytes --]

From f0d420133b58d0506173f2f1539867b2d9e7f014 Mon Sep 17 00:00:00 2001
From: nverno <noah.v.peart@gmail.com>
Date: Mon, 9 Oct 2023 22:52:38 -0700
Subject: [PATCH] Fix reset treesit--explorer-last-node

---
 lisp/treesit.el | 1 +
 1 file changed, 1 insertion(+)

diff --git a/lisp/treesit.el b/lisp/treesit.el
index 402417c6ca9..e798abac079 100644
--- a/lisp/treesit.el
+++ b/lisp/treesit.el
@@ -3192,6 +3192,7 @@ treesit-explore-mode
                          (format "*tree-sitter explorer for %s*"
                                  (buffer-name))))
             (setq-local treesit--explorer-language language)
+            (setq-local treesit--explorer-last-node nil)
             (with-current-buffer treesit--explorer-buffer
               (treesit--explorer-tree-mode)))
           (display-buffer treesit--explorer-buffer
-- 
2.34.1

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Eli Zaretskii]

> From: nvp <noah.v.peart@gmail.com>
> Date: Mon, 9 Oct 2023 23:02:48 -0700
> 
> Tags: patch
> 
> Bug: After `treesit-explorer-mode` is enabled in a buffer and its
> associated `treesit--explorer-buffer` is killed, a subsequent call
> to `treesit-explorer-mode` initially displays an empty explorer
> buffer b/c `treesit--explorer-refresh` sees old value for
> `treesit--explorer-last-node`.
> 
> * lisp/treesit.el (treesit-explorer-mode): reset
> `treesit--explorer-last-node` when `treesit--explorer-buffer` was killed
> 
> In GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
> 3.24.33, cairo version 1.16.0) of 2023-10-05 built on noah-X580VD
> Repository revision: 505c80623049d9e181918acdac8229c9a2041b1e
> Repository branch: master
> Windowing system distributor 'The X.Org Foundation', version 11.0.12101004
> System Description: Ubuntu 22.04.3 LTS

Yuan, could you please look into this?  Is the patch OK to go in, and
if so, should it be installed on emacs-29?

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Yuan Fu]

> On Oct 9, 2023, at 11:02 PM, nvp <noah.v.peart@gmail.com> wrote:
> 
> Tags: patch
> 
> 
> Bug: After `treesit-explorer-mode` is enabled in a buffer and its
> associated `treesit--explorer-buffer` is killed, a subsequent call
> to `treesit-explorer-mode` initially displays an empty explorer
> buffer b/c `treesit--explorer-refresh` sees old value for
> `treesit--explorer-last-node`.
> 
> * lisp/treesit.el (treesit-explorer-mode): reset
> `treesit--explorer-last-node` when `treesit--explorer-buffer` was killed

Hmm, I can’t reproduce what you described. Besides, treesit--explorer-last-node is only set in the source buffer, not the explorer buffer. But the patch tries to reset it for the explorer buffer. Also treesit--explorer-last-node is reset at the end of treesit-explore-mode.

We can start from reliably reproducing the bug you are seeing, and see what’s the true cause of it.

Yuan

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Yuan Fu]

> On Oct 14, 2023, at 1:17 AM, Eli Zaretskii <eliz@gnu.org> wrote:
> 
>> From: nvp <noah.v.peart@gmail.com>
>> Date: Mon, 9 Oct 2023 23:02:48 -0700
>> 
>> Tags: patch
>> 
>> Bug: After `treesit-explorer-mode` is enabled in a buffer and its
>> associated `treesit--explorer-buffer` is killed, a subsequent call
>> to `treesit-explorer-mode` initially displays an empty explorer
>> buffer b/c `treesit--explorer-refresh` sees old value for
>> `treesit--explorer-last-node`.
>> 
>> * lisp/treesit.el (treesit-explorer-mode): reset
>> `treesit--explorer-last-node` when `treesit--explorer-buffer` was killed
>> 
>> In GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
>> 3.24.33, cairo version 1.16.0) of 2023-10-05 built on noah-X580VD
>> Repository revision: 505c80623049d9e181918acdac8229c9a2041b1e
>> Repository branch: master
>> Windowing system distributor 'The X.Org Foundation', version 11.0.12101004
>> System Description: Ubuntu 22.04.3 LTS
> 
> Yuan, could you please look into this?  Is the patch OK to go in, and
> if so, should it be installed on emacs-29?

I’ll look at it, thanks for the reminder.

Yuan

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[nvp]

[-- Attachment #1: Type: text/plain, Size: 3257 bytes --]

Hi,
The patch is supposed to reset `treesit--explorer-last-node` in the source
buffer, just before the `(with-current-buffer treesit--explorer-buffer
...)`.
Upon trying to reproduce it now, I realized it's harder to reproduce than
I had thought -- sorry about that.
I noticed the bug (if it is a bug) initially when I was adding a function
to jump b/w source and explorer buffers, like the following:

(defun my-treesit-explorer-jump ()
  "Pop b/w source and explorer buffers."
  (interactive)
  (let ((buf
         (cond
          ((eq major-mode 'treesit--explorer-tree-mode)
           (when (buffer-live-p treesit--explorer-source-buffer)
             treesit--explorer-source-buffer))
          (t
           (unless (and treesit-explore-mode
                        (buffer-live-p treesit--explorer-buffer))
             ;; *** Without the reset here, the explorer buffer doesn't
             ;; get redrawn the first time, when treesit--explorer-last-node
             ;; is non-nil in the source buffer ***
             ;; (setq-local treesit--explorer-last-node nil)
             (cl-letf (((symbol-function (function completing-read))
                        (lambda (&rest _) (symbol-name (treesit-language-at
(point))))))
               (treesit-explore-mode 1)))
           treesit--explorer-buffer))))
    (pop-to-buffer buf)))

Let me give a more precise recipe to reproduce:
1. From a c++-ts-mode buffer, call `treesit-explorer-mode`, select `cpp`.
Now there should be an explorer buffer.
2. Kill the associated explorer buffer.
3. Now, back in the c++-ts-mode buffer, `treesit--explorer-last-node`
should still have a value.
4. From that c++-ts-mode buffer, call `my-treesit-explorer-jump`, and the
explorer buffer should be empty, until
switching back to the source buffer.

This seems to me to be caused by `treesit--explorer-post-command` not
running until the source
buffer is active again.


On Sat, Oct 14, 2023 at 10:08 AM Yuan Fu <casouri@gmail.com> wrote:

>
>
> > On Oct 14, 2023, at 1:17 AM, Eli Zaretskii <eliz@gnu.org> wrote:
> >
> >> From: nvp <noah.v.peart@gmail.com>
> >> Date: Mon, 9 Oct 2023 23:02:48 -0700
> >>
> >> Tags: patch
> >>
> >> Bug: After `treesit-explorer-mode` is enabled in a buffer and its
> >> associated `treesit--explorer-buffer` is killed, a subsequent call
> >> to `treesit-explorer-mode` initially displays an empty explorer
> >> buffer b/c `treesit--explorer-refresh` sees old value for
> >> `treesit--explorer-last-node`.
> >>
> >> * lisp/treesit.el (treesit-explorer-mode): reset
> >> `treesit--explorer-last-node` when `treesit--explorer-buffer` was killed
> >>
> >> In GNU Emacs 30.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version
> >> 3.24.33, cairo version 1.16.0) of 2023-10-05 built on noah-X580VD
> >> Repository revision: 505c80623049d9e181918acdac8229c9a2041b1e
> >> Repository branch: master
> >> Windowing system distributor 'The X.Org Foundation', version
> 11.0.12101004
> >> System Description: Ubuntu 22.04.3 LTS
> >
> > Yuan, could you please look into this?  Is the patch OK to go in, and
> > if so, should it be installed on emacs-29?
>
> I’ll look at it, thanks for the reminder.
>
> Yuan
>
>

[-- Attachment #2: Type: text/html, Size: 4209 bytes --]

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Yuan Fu]

[-- Attachment #1: Type: text/plain, Size: 2414 bytes --]



> On Oct 14, 2023, at 9:20 PM, nvp <noah.v.peart@gmail.com> wrote:
> 
> Hi,
> The patch is supposed to reset `treesit--explorer-last-node` in the source buffer, just before the `(with-current-buffer treesit--explorer-buffer ...)`.
> Upon trying to reproduce it now, I realized it's harder to reproduce than I had thought -- sorry about that.  
> I noticed the bug (if it is a bug) initially when I was adding a function to jump b/w source and explorer buffers, like the following:
> 
> (defun my-treesit-explorer-jump ()
>   "Pop b/w source and explorer buffers."
>   (interactive)
>   (let ((buf
>          (cond
>           ((eq major-mode 'treesit--explorer-tree-mode)
>            (when (buffer-live-p treesit--explorer-source-buffer)
>              treesit--explorer-source-buffer))
>           (t
>            (unless (and treesit-explore-mode
>                         (buffer-live-p treesit--explorer-buffer))
>              ;; *** Without the reset here, the explorer buffer doesn't
>              ;; get redrawn the first time, when treesit--explorer-last-node
>              ;; is non-nil in the source buffer ***
>              ;; (setq-local treesit--explorer-last-node nil)
>              (cl-letf (((symbol-function (function completing-read))
>                         (lambda (&rest _) (symbol-name (treesit-language-at (point))))))
>                (treesit-explore-mode 1)))
>            treesit--explorer-buffer))))
>     (pop-to-buffer buf)))
> 
> Let me give a more precise recipe to reproduce:
> 1. From a c++-ts-mode buffer, call `treesit-explorer-mode`, select `cpp`.  Now there should be an explorer buffer.
> 2. Kill the associated explorer buffer.
> 3. Now, back in the c++-ts-mode buffer, `treesit--explorer-last-node` should still have a value.
> 4. From that c++-ts-mode buffer, call `my-treesit-explorer-jump`, and the explorer buffer should be empty, until
> switching back to the source buffer.
> 
> This seems to me to be caused by `treesit--explorer-post-command` not running until the source 
> buffer is active again.

Thank you, I think I see the problem now. Could you try the below patch and see it fixes your problem? Also, sorry for the late reply, I meant to reply sooner but couldn’t find the time to figure out what exact was the cause :-) I was initially a bit confused since we already do set last-node to nil.

Yuan


[-- Attachment #2: fix-last-node.diff --]
[-- Type: application/octet-stream, Size: 912 bytes --]

diff --git a/lisp/treesit.el b/lisp/treesit.el
index c73ac9912d6..9f7e8bacd35 100644
--- a/lisp/treesit.el
+++ b/lisp/treesit.el
@@ -3198,13 +3198,13 @@ treesit-explore-mode
               (treesit--explorer-tree-mode)))
           (display-buffer treesit--explorer-buffer
                           (cons nil '((inhibit-same-window . t))))
+          (setq-local treesit--explorer-last-node nil)
           (treesit--explorer-refresh)
           ;; Set up variables and hooks.
           (add-hook 'post-command-hook
                     #'treesit--explorer-post-command 0 t)
           (add-hook 'kill-buffer-hook
                     #'treesit--explorer-kill-explorer-buffer 0 t)
-          (setq-local treesit--explorer-last-node nil)
           ;; Tell `desktop-save' to not save explorer buffers.
           (when (boundp 'desktop-modes-not-to-save)
             (unless (memq 'treesit--explorer-tree-mode

[-- Attachment #3: Type: text/plain, Size: 2 bytes --]

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[nvp]

[-- Attachment #1: Type: text/plain, Size: 3374 bytes --]

That fixes the problem!

However, the reason I initially put the reset inside the `(unless
(buffer-live-p treesit--explorer-buffer) ...)`
in `treesit-explore-mode` was b/c it looked like there was an optimization
happening in
`treesit--explorer-refresh` where it does this check

           ;; If we didn't edit the buffer nor change the top-level
           ;; node, don't redraw the whole syntax tree.
           (highlight-only (treesit-node-eq
                            top-level treesit--explorer-last-node))

I don't know if that is something you'd want to keep, but just pointing it
out in case.  I think
the initial patch works as well, but still allows that check to work when
the explorer buffer hasn't
been killed.

Thankyou!

On Wed, Oct 18, 2023 at 9:36 PM Yuan Fu <casouri@gmail.com> wrote:

>
>
> > On Oct 14, 2023, at 9:20 PM, nvp <noah.v.peart@gmail.com> wrote:
> >
> > Hi,
> > The patch is supposed to reset `treesit--explorer-last-node` in the
> source buffer, just before the `(with-current-buffer
> treesit--explorer-buffer ...)`.
> > Upon trying to reproduce it now, I realized it's harder to reproduce
> than I had thought -- sorry about that.
> > I noticed the bug (if it is a bug) initially when I was adding a
> function to jump b/w source and explorer buffers, like the following:
> >
> > (defun my-treesit-explorer-jump ()
> >   "Pop b/w source and explorer buffers."
> >   (interactive)
> >   (let ((buf
> >          (cond
> >           ((eq major-mode 'treesit--explorer-tree-mode)
> >            (when (buffer-live-p treesit--explorer-source-buffer)
> >              treesit--explorer-source-buffer))
> >           (t
> >            (unless (and treesit-explore-mode
> >                         (buffer-live-p treesit--explorer-buffer))
> >              ;; *** Without the reset here, the explorer buffer doesn't
> >              ;; get redrawn the first time, when
> treesit--explorer-last-node
> >              ;; is non-nil in the source buffer ***
> >              ;; (setq-local treesit--explorer-last-node nil)
> >              (cl-letf (((symbol-function (function completing-read))
> >                         (lambda (&rest _) (symbol-name
> (treesit-language-at (point))))))
> >                (treesit-explore-mode 1)))
> >            treesit--explorer-buffer))))
> >     (pop-to-buffer buf)))
> >
> > Let me give a more precise recipe to reproduce:
> > 1. From a c++-ts-mode buffer, call `treesit-explorer-mode`, select
> `cpp`.  Now there should be an explorer buffer.
> > 2. Kill the associated explorer buffer.
> > 3. Now, back in the c++-ts-mode buffer, `treesit--explorer-last-node`
> should still have a value.
> > 4. From that c++-ts-mode buffer, call `my-treesit-explorer-jump`, and
> the explorer buffer should be empty, until
> > switching back to the source buffer.
> >
> > This seems to me to be caused by `treesit--explorer-post-command` not
> running until the source
> > buffer is active again.
>
> Thank you, I think I see the problem now. Could you try the below patch
> and see it fixes your problem? Also, sorry for the late reply, I meant to
> reply sooner but couldn’t find the time to figure out what exact was the
> cause :-) I was initially a bit confused since we already do set last-node
> to nil.
>
> Yuan
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 4223 bytes --]

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Yuan Fu]

> On Oct 20, 2023, at 2:22 PM, nvp <noah.v.peart@gmail.com> wrote:
> 
> That fixes the problem!
> 
> However, the reason I initially put the reset inside the `(unless (buffer-live-p treesit--explorer-buffer) ...)`
> in `treesit-explore-mode` was b/c it looked like there was an optimization happening in 
> `treesit--explorer-refresh` where it does this check 
> 
>            ;; If we didn't edit the buffer nor change the top-level
>            ;; node, don't redraw the whole syntax tree.
>            (highlight-only (treesit-node-eq
>                             top-level treesit--explorer-last-node))
> 
> I don't know if that is something you'd want to keep, but just pointing it out in case.  I think
> the initial patch works as well, but still allows that check to work when the explorer buffer hasn't 
> been killed.

Oh that’s fine, treesit-explore-mode always wipes everything and start from a clean slate. That optimization is for when the user moves point in the source buffer when explore-mode is on. If you’d like to send a patch that does roughly what I did in the patch I sent, I’d love to merge it. Otherwise I can fix it myself, too.

The initial patch could be a bit confusing to the readers since it sets last-node twice, and it’s not clear why.

Thanks,
Yuan

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[nvp]

[-- Attachment #1: Type: text/plain, Size: 1564 bytes --]

Ok that makes sense, thanks for clearing that up for me, your fix looks
good. I'm loving this package!

On Sat, Oct 21, 2023 at 11:33 AM Yuan Fu <casouri@gmail.com> wrote:

>
>
> > On Oct 20, 2023, at 2:22 PM, nvp <noah.v.peart@gmail.com> wrote:
> >
> > That fixes the problem!
> >
> > However, the reason I initially put the reset inside the `(unless
> (buffer-live-p treesit--explorer-buffer) ...)`
> > in `treesit-explore-mode` was b/c it looked like there was an
> optimization happening in
> > `treesit--explorer-refresh` where it does this check
> >
> >            ;; If we didn't edit the buffer nor change the top-level
> >            ;; node, don't redraw the whole syntax tree.
> >            (highlight-only (treesit-node-eq
> >                             top-level treesit--explorer-last-node))
> >
> > I don't know if that is something you'd want to keep, but just pointing
> it out in case.  I think
> > the initial patch works as well, but still allows that check to work
> when the explorer buffer hasn't
> > been killed.
>
> Oh that’s fine, treesit-explore-mode always wipes everything and start
> from a clean slate. That optimization is for when the user moves point in
> the source buffer when explore-mode is on. If you’d like to send a patch
> that does roughly what I did in the patch I sent, I’d love to merge it.
> Otherwise I can fix it myself, too.
>
> The initial patch could be a bit confusing to the readers since it sets
> last-node twice, and it’s not clear why.
>
> Thanks,
> Yuan

[-- Attachment #2: Type: text/html, Size: 2039 bytes --]

bug#66431: [PATCH] Fix reset treesit--explorer-last-node when explorer buffer was killed

[Yuan Fu]

> On Oct 21, 2023, at 5:40 PM, nvp <noah.v.peart@gmail.com> wrote:
> 
> Ok that makes sense, thanks for clearing that up for me, your fix looks good. I'm loving this package!

Thanks, I pushed the fix to emacs-29. I’m glad you are finding it useful!

Yuan